While writing my previous blog, When Will AI Come After YOUR Job? (Asking for a Friend…), I realized how much things have progressed in AI space since 2019, which was when I posted my piece on What the FAQ are AI, ANNs, ML, DL, and DNNs?
There are all sorts of new AI-based concepts and technologies bouncing around, and it can be hard to wrap your head around them, so this blog provides a brief overview to at least give you a fighting chance.
Perceptive AI
Perceptive AI refers to systems that can sense, interpret, and understand the physical world using data from sensors, especially vision, audio, and other real-world inputs. We can think of it as AI that can answer questions like “What am I looking at?”, “What am I hearing?”, and “What’s happening around me right now?”
Typical capabilities include object detection (cars, people, cats, etc.), image classification, facial recognition, speech recognition, and sensor fusion. In preceptive AI systems, sensor fusion is the process of combining data from multiple sensors to create a more accurate and reliable understanding of the real world than any single sensor could provide alone. If Perceptive AI gives machines their senses, then sensor fusion is what stops them from jumping to conclusions. It’s the difference between “I heard a noise” and “I heard a noise, saw something move, and yes—that’s definitely the stupid cat again.”
Enhancive AI
Enhancive AI refers to systems that improve, restore, refine, or augment existing data, content, or processes. Instead of asking “What is this?” (perceptive AI), enhancive AI asks “How can I make this better?”
To put this another way, enhancive AI doesn’t create something new; it takes what you already have and makes it better. Typical examples include upscaling low-resolution images, removing noise from audio signals, enhancing video quality and frame rates, cleaning up sensor data, and optimizing workflows and system performance. In all these cases, the input already exists; the AI’s job is to polish, sharpen, or otherwise improve its quality or usefulness.
For example, an image enhancement system might take a blurry, low-resolution photo and transform it into a crisp, detailed image. Similarly, an audio enhancement system might remove background noise and clarify speech, while a signal-processing system might extract meaningful information from noisy sensor readings.
Assistive AI
Assistive AI sits somewhere between “helpful sidekick” and “turbocharged tool.” Rather than doing the job for you, it helps you do a better job yourself—offering suggestions, highlighting issues, and nudging you in the right direction. Think spell checkers, grammar checkers, code assistants, and recommendation engines. In all these cases, the human remains firmly in charge; the AI just makes them faster, sharper, and (one hopes) slightly less error-prone.
Generative AI
Generative AI refers to systems that can create new content—text, images, audio, video, code, and more—based on patterns learned from vast amounts of data. Instead of answering “What is this?”, Generative AI answers questions like “Write me a story about…”, “Draw me a picture of…”, and “Generate some code that does…”
Typical examples include large language models (LLMs), image generators, music generators, and code-generation tools. The key thing to remember is that Generative AI doesn’t “know” things the way humans do. Instead, it has learned statistical patterns and uses those patterns to produce outputs that are often remarkably convincing (and occasionally hilariously wrong).
Agentic AI
Agentic AI refers to autonomous systems that can act independently to achieve complex, multi-step goals with minimal human oversight. Unlike passive, generative AI systems that simply respond to prompts, agentic systems ask, “What needs to be done—and how do I get it done?”
These systems combine reasoning, planning, and the use of external tools to proactively execute full workflows. Instead of generating a single response, they can break a task into steps, decide what actions to take, call tools or application programming interfaces (APIs), evaluate results, and adjust their approach as needed
For example, an agentic system might research a topic, generate a report, refine it, and send it to the appropriate people—all with minimal human intervention. Of course, “autonomous” doesn’t mean “infallible.” Left unsupervised, some agents can wander off course like an overconfident intern with admin privileges.
A common analogy is to think of Agentic AI as the conductor of an orchestra, with individual AI agents acting as the musicians. The conductor interprets the goal, decides who should do what and when, and keeps everything in sync, while each agent performs a specific task—researching, coding, writing, or analyzing. In more advanced systems, the conductor can even “bring in” new musicians as needed, spawning additional agents to handle different parts of the job. The result is a coordinated ensemble capable of tackling complex, multi-step problems that would overwhelm any single player.
Of course, as with any orchestra, things can occasionally go awry. Left unchecked, one enthusiastic violinist (agent) may decide to improvise, while the percussion section (another agent) enthusiastically follows along, and before you know it, the whole performance has taken on a life of its own, and instead of listening to a stately classical piece, we find ourselves tap-dancing to a polka.
Physical AI
Before we dive into Physical AI, it’s worth briefly touching on two other terms: the cloud and the edge. The cloud refers to large, centralized data centers packed with computing power. This is where AI models are typically trained and where heavy-duty processing can take place. The edge refers to devices out in the real world—things like cameras, drones, robots, cars, and industrial machines—where decisions often need to be made quickly, locally, and sometimes without a reliable internet connection.
Now, with that under our belts… physical AI refers to systems that interact directly with the physical world through machines and devices. Instead of just perceiving, enhancing, generating, or deciding, Physical AI asks, “What action should I take—and how do I physically carry it out?”
These systems bring together many of the capabilities we’ve discussed so far—perception, enhancement, reasoning, generation, and agency—allowing them to interact with and influence the physical world:
- Perceptive AI to understand the environment.
- Enhancive AI to clean up and refine incoming data.
- Generative AI to model possibilities, predict outcomes, or create plans.
- Agentic AI to make decisions and coordinate actions.
"Ah ha!" I hear you cry, "You forgot to include assistive AI in this list." Not so. I didn't forget it; I omitted it intentionally. Physical AI is about embodiment plus action in the real world, while assistive AI is more about human interaction and augmentation. So while assistive AI is often present in physical AI systems (interfaces, guidance, feedback), it isn't foundational in the same way as perception, generation, or agency.
Typical examples of physical AI include robots in factories and warehouses, autonomous vehicles and drones, smart home, office, and city devices, medical robots, prosthetics, and rehabilitation devices. In many cases, the “brains” may live partly in the cloud (for training and high-level reasoning) and partly at the edge (for real-time decision-making and control). This split is important because when you’re about to bump into a wall, you don’t want to wait for decisions to make a round trip to a data center on the other side of the planet.
Human in the Loop (HITL)
As AI systems become more capable—and more autonomous—it’s important to remember that, in many cases, humans are still part of the process. This concept is often referred to as Human in the Loop (HITL). This refers to systems in which a human provides oversight, guidance, or final approval, especially when the stakes are high or the consequences of getting things wrong are… undesirable.
For example, a medical AI might suggest a diagnosis, but a doctor makes the final call; an autonomous vehicle might handle most driving tasks, but a human can take over when needed; an agentic system might draft a report, but a human reviews it before it’s sent. In all these cases, the AI does much of the heavy lifting, but the human remains the ultimate arbiter of correctness and responsibility.
Human in the Loop is what keeps increasingly capable AI systems aligned with human judgment, values, and common sense (such as it is). It’s also what prevents our previously mentioned orchestra from marching off into the nearest town square and launching into an impromptu polka recital at entirely the wrong moment.
Vibe Coding
Finally, we come to something rather different: Vibe Coding. Unlike the other topics we discussed earlier, Vibe Coding isn’t a type of AI—it’s a way of working with AI. Vibe Coding is a software development approach in which, instead of writing code line by line, developers use natural language prompts to guide AI systems—often agentic ones—to generate, debug, and refine applications. Coined in early 2025, the term emphasizes high-level intent and rapid iteration (“vibes”) over manual syntax.
Instead of saying, “Here is a precisely defined specification,” you say something more like, “I’m trying to build something like this… let’s see where this goes.” You describe the vibe of what you want, and the AI helps you explore, refine, and implement it. In many ways, Vibe Coding sits on top of everything we’ve discussed:
- Perceptive AI helps systems understand the world.
- Enhancive AI improves the inputs.
- Assistive AI supports the user.
- Generative AI creates possibilities.
- Agentic AI organizes and executes tasks.
- Physical AI makes things happen in the real world.
Vibe Coding is how we orchestrate all of the above—using conversation instead of code. With Vibe Coding, you might describe an application in plain English, have an AI generate the initial code, ask it to fix bugs, add features, or optimize performance, let agentic systems test and refine the result, and iterate until the solution “feels right.”
You’re no longer programming in the traditional sense—you’re guiding. If traditional programming is like carefully writing sheet music note by note, then Vibe Coding is more like standing in front of the orchestra and saying, “Let’s try something a bit more jazzy,” and seeing what happens.
In Conclusion
If we step back and squint a little, we can see that these different flavors of AI form a kind of continuum. Enhancive AI cleans up the inputs, perceptive AI helps machines understand what’s going on, generative AI creates new possibilities, agentic AI decides what to do, and physical AI makes it happen in the real world. Threaded through all of this is assistive AI, which keeps humans in the loop—guiding, suggesting, and collaborating along the way. What started as simple pattern recognition has blossomed into systems that can perceive, reason, create, and act, often with us working hand-in-hand (well, hand-in-algorithm) alongside them.
And now, with Vibe Coding sitting on top of it all, we’re changing the way we interact with these systems. Instead of laboriously specifying every detail, we’re learning to guide, nudge, and collaborate, sometimes with surprising (and occasionally slightly alarming) results. The future of AI may not be about telling machines exactly what to do, but about learning how to work with them effectively… and perhaps developing a good ear for when the music starts drifting toward polka.













Nice overview, summary Clive!
Thanks Jesse — the funny thing is that someone just sent me an email with a great description of the process of creating a model, training it in floating-point, then converting it into fixed-point for inferencing… and I thought “That’s a great description — I wonder where it came from” and then I realized it was from my own earlier column What the FAQ are AI, ANNs, ML, DL, and DNNs? ROFLOL